:: Re: [Dng] Fwd: Re: eudev (Was: us…
Startseite
Nachricht löschen
Nachricht beantworten
Autor: Matteo Panella
Datum:  
To: dng
Betreff: Re: [Dng] Fwd: Re: eudev (Was: useful blog post)
I have to perform a bit of quoting surgery so I won't send out a barrage
of replies, please forgive me for any error :)

On 29/11/2014 23:37, chris wrote:
> If you're specifically NOT using systemd why would you want systemd-udev ?
> ( or for that matter anything systemd-* )


It's not a matter of what I want (frankly, I'd prefer moving off
systemd-udev Real Soon Now™, given that the upstream explicitly stated
that once kdbus is merged in mainline it's game over for udev without
systemd as PID1), it's a matter of not breaking existing packages which
depend on libudev/libgudev alone, not other parts of the systemd bandwagon.

> On Gentoo I use eudev and everything compiles happily


Been there, done that. API compatibility is already there (taking for
granted that we're talking about "equivalent" versions of eudev and
systemd-udev), ABI compatibility OTOH is a whole different beast. If the
latter cannot be guaranteed, we could only rely on the former, thus
forcing a full rebuild of *every single rdep*. That means burning a lot
of CPU time (not to mention archive space), so it should be avoided at
all costs if possible.

(though I have to say that in my humble experience I haven't faced
dynamic linking breakage when switching to eudev - but that says very
little about overall ABI breakage)

[Ruben Safir]
> No, it is ironic only that one might confuse the problem between freedom
> of software and freedom of users.


Fair enough, I was mislead by the "default" and missed altogether the
"replaced" that followed it. In that case, the only thing that should be
sorted out is ABI compatibility: if that holds, it's just a matter of
producing a set of udev packages with the same names as those produced
by systemd. devuan-baseconf then would take care of giving them a higher
priority over those of the main Debian archive, thus effectively
replacing systemd-udev with eudev.

> That is going to be the theme of the next 5 years if you are serious
> about a real fork.


I hope that eudev will be the theme of the next few _weeks_, not years.

Kind regards,
--
Matteo Panella